How we score
Small Sheffield businesses get a sustainability profile. Larger businesses get one ESG score out of 10. This page explains both, and where they can be wrong.
The directory right now
- Businesses with a sustainability profile
- 13
- Businesses with an ESG score
- 0
13 Sheffield businesses are listed.
Counted from the live directory, refreshed hourly.
Why small businesses get a profile, not a score
A small business usually leaves a small digital footprint: a simple website, a few listings, little written down about its practices. A number built from that would mostly measure how much a business has published, so it would not be reliable. We do not show a score, a rank or a grade for small businesses.
Instead, the same kind of AI-assisted web research looks for public information about the business, and we show what it found as a plain list:
- What we found: positives such as strengths, green practices, sustainability initiatives, local values and community involvement.
We show how many public sources the research listed. That is a count of links, not a check that each claim is true. The profile may be incomplete or out of date, and a business with little online shows “Not much public information found yet” rather than empty boxes. A short list does not mean a business does badly; it can mean little is written down.
How larger businesses are scored
Larger businesses (business type Enterprise) are the ones that get an ESG score. The rest of this page explains how that number is produced. What the score is, and is not: it comes from AI-assisted web research. For each business, an AI model (Google Gemini, with Google Search) looks for public information about it and writes down specific environmental, social and governance practices it finds, each with a source. A script then turns that written research into numbers.
- It is not an audit, an inspection or a certification.
- The process does not include visiting the business or checking its claims on site.
- The scoring script does not open the sources or judge whether a claim is true. It counts how much of the researched text covers the topics listed below.
- A high score means a lot of relevant public evidence was found. A low score can mean little was found, not that the business does badly.
The three areas and their weights
Each area is scored from 0 to 10. The overall score is the weighted total: environmental 35%, social 45% and governance 20%. Social carries the most weight.
Within an area, the script looks for the words and phrases below in the research text. That text is the descriptions the AI wrote, every evidence item, and any certifications, initiatives and strengths it listed. A phrase counts once for variety, and repeated mentions add a small extra.
Environmental35%
carbon, emissions, renewable energy, solar, wind, sustainability, environmental, climate, green, eco-friendly, carbon neutral, waste reduction, recycling, circular economy, biodiversity.
Social45%
diversity, inclusion, employee, community, safety, health, human rights, fair trade, social impact, education, training, wellbeing, gender equality, local employment, charitable.
Governance20%
transparency, governance, ethics, compliance, board, audit, disclosure, stakeholder, accountability, integrity, risk management, anti-corruption, data protection, responsible.
How an area score is calculated
- Variety. The share of the area’s keywords that appear at least once, multiplied by 10. Finding 6 of 15 keywords gives 4.0.
- Repetition. 0.1 for every mention, up to 2.0 extra.
- Evidence bonus. Each evidence item adds points by its own confidence label: 0.5 for high, 0.3 for medium, 0.1 for low, up to 1.5 extra.
- The area score is the total, capped at 10 and rounded to one decimal place.
Because the script reads words, not meaning, a certification such as B Corp or ISO 14001 earns nothing by itself. It only counts if its name contains one of the keywords, or the AI’s write-up around it does.
Evidence and confidence
Evidence
An evidence item is one specific practice the AI reports for a business, such as a recycling scheme or a charity partnership, filed under environmental, social or governance, with the place it was found. The AI searches the business’s own site, review sites, social media, local news, forums, charity pages and registers such as Companies House, the Charity Commission, the B Corp directory and the Living Wage Foundation list. Each item carries a high, medium or low label, which the AI assigns. It is told to use high only when it has a direct source link.
Confidence
Confidence is a number from 0 to 1 stored alongside the score. It does not change the score. It is a guide to how much to trust it. It starts from the research quality the AI reports for its own search: 0.90 for excellent, 0.75 for good, 0.55 for limited and 0.35 for very limited. Each evidence item then adds 0.01, up to 0.10 in total, and the result is capped at 1.
Confidence therefore reflects how much the AI says it found, and how many evidence items it listed. It is not an independent check of the score.
What the bands mean
The scoring script produces only the 0 to 10 number. The five bands below are labels this site puts on that number. They apply to larger businesses only.
- Exceptional9.0 - 10.0
- Industry-leading sustainability practices across all areas.
- Excellent7.5 - 8.9
- Strong commitment to sustainability with measurable impact.
- Good6.0 - 7.4
- Solid sustainability practices with ongoing improvement.
- Developing4.5 - 5.9
- Making progress on sustainability with room to grow.
- EmergingBelow 4.5
- Early in its sustainability journey, with basic practices in place.
“Not yet scored”
It appears on a larger business that has no ESG score recorded yet. It is not a low score and it is not a judgement. We show it rather than guess. Small businesses never show it, because they get a profile instead.
Think a profile or score is wrong?
Because the research is automated, it can miss things or pick up the wrong business. If you own or work for a business and something is missing or incorrect, email us with the business name and what is wrong.